Browser-act is a browser automation CLI built for AI agents. It helps agents open pages, read page state, click, type, extract data, manage sessions, bypass common blocking, and hand control to a human when automation gets stuck.Documentation Index
Fetch the complete documentation index at: https://docs.browseract.com/llms.txt
Use this file to discover all available pages before exploring further.
Start here
Install
Install the Browser-act CLI, add the entry Skill to your agent, and verify that the runtime is available.
Quick Start
Run your first extraction or full browser automation loop with Browser-act.
Command Reference
Look up browser, session, stealth, CAPTCHA, network, and system commands.
Why Browser-act
AI agents need a browser that goes beyond standard automation tools.Bypass blocking
Use stealth fingerprints, TLS rotation, proxies, CAPTCHA solving, and remote handoff when a site challenges automation.
Choose the right browser
Use local Chrome state, direct Chrome control, privacy-focused stealth sessions, or fixed-identity stealth browsers.
Run concurrent work
Keep browsers, sessions, cookies, fingerprints, and proxies isolated so multiple tasks do not interfere with each other.
Work in model-friendly state
Read compact indexed state, act with commands such as
click 3, and reuse browsers by semantic descriptions.Core capabilities
| Capability | What it helps with |
|---|---|
| Anti-detection and blocking | Reach pages that block basic fetchers or standard automation |
| Better headless mode | Run silently without giving up stealth or human handoff |
| Remote assist | Let a human complete 2FA, CAPTCHA, or judgment-heavy steps from any device |
| Browser modes | Pick between Chrome, chrome-direct, stealth privacy, and stealth fixed identity |
| Concurrency and isolation | Run many agents, tasks, accounts, and sessions without state leaks |
| Skill Forge | Turn a one-time site exploration into a reusable scraping skill |
Compatibility
| Area | Support |
|---|---|
| Operating systems | Windows, macOS, Linux |
| Agent environments | Claude Code, Cursor, VS Code, OpenCode, OpenClaw, Codex, Gemini CLI, and other agents that can run shell commands and load skills |
| Runtime objects | browser, session, profile, stealth browser, network capture, HAR, cookies |